B R A I N R E S E A R C H 1 0 9 4 ( 2 0 0 6 ) 1 7 9 – 1 9 1

ava i l ab l e a t www.sc i enced i rec t . com

www.e l sev i e r. com/ l oca te /b ra in res

Research Report

The perception of musical phrase structure: A cross-culturalERP study

Yun Nana,b,c, Thomas R. Knöschea,c,⁎, Angela D. Friedericia

aMax Planck Institute for Human Cognitive and Brain Sciences, Stephanstrasse 1A. 04103 Leipzig, GermanybNational Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University, Beijing, ChinacKey Laboratory of Mental Health, Institute of Psychology, Chinese Academy of Sciences, Beijing, China

A R T I C L E I N F O

⁎ Corresponding author. Max Planck Institute+49 3425 887511.

E-mail address: knoesche@cbs.mpg.de (T

0006-8993/$ – see front matter © 2006 Elsevidoi:10.1016/j.brainres.2006.03.115

A B S T R A C T

Article history:Accepted 28 March 2006Available online 18 May 2006

Electroencephalography (EEG) was used in a cross-culturalmusic study investigating phraseboundary perception. Chinese and German musicians performed a cultural categorizationtask under Chinese and Western music listening conditions. Western music was the majorsubject for both groups of musicians, while Chinese music was familiar to Chinese subjectsonly. By manipulating the presence of pauses between two phrases in the biphrasalmelodies, EEG correlates for the perception of phrase boundaries were found in both groupsunder both music listening conditions. Between 450 and 600 ms, the music CPS (closurepositive shift), which had been found in earlier studies with a false tone detection task, wasreplicated for the more global categorization task and for all combinations of subject groupand musical style. At short latencies (100 and 450 ms post phrase boundary offset), EEGcorrelates varied as a function of musical styles and subject group. Both bottom–up (styleproperties of themusic) and top–down (acculturation of the subjects) information interactedduring this early processing stage.

© 2006 Elsevier B.V. All rights reserved.

Keywords:Music perceptionPhrase structureCultural differenceEvent-related potential

1. Introduction

Music is a culturally specific phenomenon, which is certainlydetermined by ethnic background, social environment andtraditions. However, the question to what extent the ability tounderstand and appreciate music is culturally dependent isstill controversial. Ethnomusicologists usually assume thatextensive experience of music within its cultural context isessential for its basic structural principles to be understood(Blacking, 1973). Lynch and Eilers (1992) explored musicaltuning perception in infancy and adulthood and found thatadult's perception of musical tuning was culturally specific,and infants between 6 and 12 months of age started to show

for Human Cognitive an

.R. Knösche).

er B.V. All rights reserved

the same culturally specific performance as adults did. In thesame line, Drake and El Heni (2003) proved that acculturation,i.e., the passive exposure to a particular type of music sincebirth, influenced the perception and cognition ofmusic. On theother hand, some psychological studies have proved theexistence of universal, not culturally dependent, aspects ofmusic perception (Castellano et al., 1984; Krumhansl, 1995,2000). Drake and Bertrand (2001) reviewed a number of rhythmperception and production studies and concluded that someaspects of temporal processing inmusicmight be universal, inthe sense that they function in a similarmanner irrespective ofan individual's cultural exposure and experience. Likewise, ina study of the cross-cultural emotion perception in music,

d Brain Sciences, Stephanstrasse 1A. 04103 Leipzig, Germany. Fax:

.

180 B R A I N R E S E A R C H 1 0 9 4 ( 2 0 0 6 ) 1 7 9 – 1 9 1

Balkwill and Thompson (1999) reported that Western listenerswere sensitive to musically expressed emotions in an unfa-miliar tonal system (Hindustani ragas). The authors suggestedthat the perception of emotion in unfamiliar music isfacilitated by psychophysical but not cultural cues. Further-more, there is evidence from developmental studies suggest-ing that human musicality is present from birth, and that itsfoundations are not culturally constructed. Smith and Wil-liams (1999) conducted an experiment about children's artisticresponses to musical intervals. They asked children fromdifferent cultural backgrounds to draw pictures when theywere hearing music intervals. The results suggested that theperception of musical meaning is a universal rather thanculturally based phenomenon.

As two main representations of auditory information,music and language hold similar acoustic features. Therelationship between music and language has recently beeninvestigated by a number of neurophysiological studies atdifferent processing levels. Koelsch et al. (2004) found thatmusic, similar to language, could prime the semantic proces-sing in the humanbrain. At the syntactic level, incongruities inmusic elicit brain responses similar to the ones observed inlanguage experiments (Besson, 1998;Maess et al., 2001; Patel etal., 1998). In addition,Maess and co-workers (Maess et al., 2001)used magnetoencephalography (MEG) to demonstrate thatBroca's area, held responsible for the processing of syntaxduring language processing, was also involved in the earlyprocessing of music–syntactic incongruities.

Music and language are both composed of sequentialevents unfolding in time. This property renders segmentalinformation, including phrase structure, very important in theon-line perception in both domains. The phrase perceptionprocess is normally induced by certain phrase boundary cues,like the insertion of a pause and the lengthening of the prefinalsyllable or musical note. In language, phrasing has beeninvestigated in behavioral (e.g., Warren et al., 1995) andneurophysiological studies (e.g., Steinhauer et al., 1999).Results from an event-related brain potential (ERP) study(Steinhauer et al., 1999) revealed a positive shift (ClosurePositive Shift, CPS) in a close relationship with intonationalphrase boundaries,whichwas interpreted to reflect the closingprocess of the prior intonational phrase. Later studies(Pannekamp et al., 2005; Steinhauer and Friederici, 2001)added more evidence to the notion that indeed prosodicrather than syntactic cues give rise to the CPS. Mostinterestingly, the investigation of phrase perception inmusic (Knösche et al., 2005) yielded EEG and MEG correlatesfor the perception of musical phrase boundaries, whichshowed a similar timing and topography as the CPS found inlanguage studies. In a follow-up study (Neuhaus et al., 2006),the length of the pause and the length of the phrase-finaltone before the pause were identified as the most importantmarkers for the perception of the phrase boundaries. Thisresult is in line with a recent behavioral study by Franklandand Cohen (2004), who found the note-to-note distances tobe the most important phrase boundary cues.

Developed through exposure to particular culture-basedenvironments, the processing of music and language is bothculturally tuned. From cross-linguistic studies, there isaccumulating evidence demonstrating that the same brain

regions in the left hemisphere are activated for the processingof semantic and syntactic aspects (across different languages)when subjects process their native language (Gandour et al.,2004; Jacobsen et al., 2004; Klein et al., 2001; Schlosser et al.,1998; for a review, see Friederici and Rueschemeyer, in press).In contrast, cultural differences in the neural basis of musicperception, though currently attracting much interest, haveyet to be exploredmore extensively. There are, however, someaccounts for cultural specificity of the brain processesassociated with music perception (see, e.g., Neuhaus, 2003).As suggested by Arikan et al. (1999) for Turkish subjects,cultural effects could be shown when auditory oddball stimuliwere superimposed on white noise, silent background, musicplayed in familiar (ney) or unfamiliar instrument (violoncello).Larger P3 amplitudes were observed in the ney backgroundcondition as compared to violoncello one. This ERP result wasinterpreted as a facilitated attention allocation and memoryupdating processes when hearing music of a familiar style. Ina musicogenic epilepsy case study, Genc et al. (2001) showedthat only Turkishmusic was able to trigger seizures in this 48-year-old Turkish female but not other musical styles, e.g.,Western music. Those two studies suggest distinct neuralsubstrates engaged in the processing of culturally familiarmusic in comparison to unfamiliar music. In contrast, afunctionalmagnetic resonance imaging (fMRI) study (Morrisonet al., 2003) failed to prove any activation difference based oncultural familiarity of music either in musical experts or innaive subjects. The authors thus argued that listening toculturally different music might activate quite similar neuralregions. Whether this result is due to the low temporalresolution (500 ms to 1000 ms) not allowing a differentiationof early auditory processes as investigated by a standardoddball paradigm and the indirect measures of brain activityby fMRI, or indeeddue to a largely commonneural substrate forthe processing of culturally familiar and unfamiliar music, isnot yet clarified.

The aim of the present study is to determine whethermusical experts from different culture backgrounds differfrom each other in their neural activation associated withcultural familiar/unfamiliarmusic processes. By using EEG, weaimed at revealing the interactions between cultural variablesand music processes by highlighting the time course ofprocessing. In particular, we investigated the perception ofsegmental information, i.e., phrasing. As mentioned above,the CPS has been proven to be a marker for the phraseperception in bothmusic and language domains. Based on theresults from previous comparative studies betweenmusic andlanguage (see above and see review: Patel, 2003), it seemsreasonable to predict a universal phrase perception mecha-nism. We used Chinese and Western musical excerpts for thecurrent study. Compared to seven major or minor heptatonicscales for Western music, Chinese music mainly uses fivepentatonic basic scales. A simple impression of how thesescales sound like could be obtained by composing and playinga piece that can be performed by using only the black keys ofthe piano. By comparing brain activation differences ofGerman and Chinese musicians in Western and Chinesemusic listening conditions, we intend to elucidate whetherphrase perception is universal in music processing and towhat extent it is modified by cultural differences.

Table 1 – Behavioral results: statistics for the percentagesof correct answers

(a) Outcome of the three-way ANOVA: COND*STYLE*GROUP.The degrees of freedom for all F values are (1, 20). CONDstands for experimental condition (phrased, unphrased);STYLE indicates style of music (Chinese, Western); GROUPrefers to subject group (Chinese, German). Only P valuesbelow 0.1 are shown

Correct answers percent

COND No significanceSTYLE*GROUP F=12.0, P<0.01COND*GROUP No significanceCOND*STYLE No significanceCOND*STYLE*GROUP F=7.8, P<0.05GROUP F=11.6, P<0.01

(b) Pair-wise comparison for STYLE*GROUP interaction (standarddeviation in parenthesis)

P value Difference(1st item

181B R A I N R E S E A R C H 1 0 9 4 ( 2 0 0 6 ) 1 7 9 – 1 9 1

2. Results

2.1. Behavioral results

The behavioral results are depicted in Fig. 1. The statistics arelisted in Table 1. Clearly, both groups perform better with theirown native style of music (interaction GROUP × STYLE;significant pair-wise differences between styles in eachgroup; see Table 1b and Fig. 1A). Moreover, the Chinese showa higher performance than the Germans for Chinese tradi-tional music, while there was no difference between thegroups for Western music (main effect GROUP, significantpair-wise difference between groups for Chinese music, butnot for Western music).

Additionally, it seems that for listening to culturallyunfamiliar music (Chinese musicians listening to Westernmusic and German musicians listening to Chinese music),both groups performed better for unphrased as compared tophrasedversions, as hintedby a significant interactionCOND×

minus 2nd one)

German musicians: P<0.05 –4.625 (1.62)Chinese musicvs. Western music

Chinese musicians: P<0.05 2.10 (0.843)Chinese musicvs. Western music

Chinese music: P<0.001 7.225 (1.316)Chinese musiciansvs. German musicians

Western music: No significance 0.5 (1.649)Chinese musiciansvs. German musicians

(c) Pair-wise comparison between unphrased and phrasedconditions for STYLE*GROUP*COND interaction. The table refersto the difference between phrased and unphrased items asdependent variable

P value Difference(1st item

minus 2nd one)

German musicians: P<0.1 1.250 (0.653)Chinese musicvs. Western music

Chinese musicians: No significance −0.667 (0.899)Chinese musicvs. Western music

Chinese music: No significance −0.8 (0.467)Chinese musiciansvs. German musicians

Western music: P<0.1 1.40 (0.703)Chinese musiciansvs. German musicians

Fig. 1 – Behavioral results: percentage of correct answers. Fordifferent music styles over two groups of musicians, thediagrams illustrate (A) the interaction between STYLE(Chinese vs. Western music)*GROUP (Chinese vs. Germansubjects) and (B) the interaction between COND (phrased vs.unphrased melodies)*STYLE*GROUP. Error bars indicatestandard deviations.

GROUP × STYLE. However, the differences were small and therespective pair-wise comparisons only marginally significant(see Table 1c and Fig. 1B).

2.2. Electrophysiological results

2.2.1. P3 experimentThe ERP traces for standard and deviant tones in differentmusician groups are shown in Fig. 2A. For both subject

Fig. 2 – Grand average ERPs for the P3 experiment. (A) Time courses: the grey lines refer to Chinese musicians, black lines toGerman musicians; the solid lines stand for deviant tones, while dotted lines stand for standard tones. The analyzed timewindows are labeled as follows: I (0–130 ms), II (130–240 ms), and III (240–450 ms). (B) Topographic maps: the ERPs wereintegrated over the time window III (300–450 ms). The maps are viewed from above, the nose is pointing upwards.

182 B R A I N R E S E A R C H 1 0 9 4 ( 2 0 0 6 ) 1 7 9 – 1 9 1

Fig. 3 – Time courses of the grand average ERPs for themain experiment: The grey lines indicate Chinesemusicians, black linesfor German musicians; the solid lines indicate phrased melodies listening, dotted lines for unphrased melodies listening. Thechosen time windows are labeled as follows: I (40–100 ms), II (100–300 ms), III (300–450 ms), and IV (450–600 ms). Note that thebaseline for all conditions was set prior to the onset of the phrase boundary in the phrased condition and its respective point inthe unphrased condition (not the pretrigger interval shown in the diagrams). (A) Chinese music. (B) Western music.

183B R A I N R E S E A R C H 1 0 9 4 ( 2 0 0 6 ) 1 7 9 – 1 9 1

Table 2 – Results of statistical analysis of the main experiment over all the subjects (mean amplitude in time windows)

Analysis window I: 40–100 ms (N1) II: 100–300 ms (P2) III: 300–450 ms IV: 450–600 ms (CPS)

COND F = 11.8, P < 0.003 F = 52.7, P < 0.001 F = 25.7, P < 0.001 F = 7.8, P < 0.05COND*STYLE No significance F = 6.5, P < 0.05 F = 5.4, P < 0.05 F = 6.1, P < 0.05COND*GROUP No significance F = 4.7, P < 0.05 No significance No significanceCOND*STYLE*GROUP No significance F = 4.2, p = 0.055 F* = 4.9, P < 0.01 No significance

COND stands for experimental condition (phrased, unphrased); STYLE refers to style of music (Chinese, Western); GROUP indicates subject group(Chinese, German). Only P values below 0.1 are shown. Only effects involving the factor COND are considered. The degrees of freedom for all Fvalues are (1, 20). F* indicates the significance involving the interactions with recording channel (51 levels), thus the respective degrees offreedom are (50, 1000).

Fig. 4 – Mean amplitudes of the ERP components for themainexperiment: graphs show the amplitude differences(averaged over all 51 channels) between phrased andunphrased conditions ( phrase/ unphrased) over all the fourcombinations of music styles and musicians groups, for thetime windows II (100–300 ms), III (300–450 ms), and IV(450–600 ms).

184 B R A I N R E S E A R C H 1 0 9 4 ( 2 0 0 6 ) 1 7 9 – 1 9 1

groups and stimulus conditions, we observe a biphasic N1-P2 pattern, peaking at about 100 and 200 ms after stimulusonset, respectively. While the N1 shows no differencesbetween subject groups or between stimulus conditions,the amplitude of the P2 seems to be slightly reduced forthe deviant stimuli at some electrodes, in particular forWestern subjects (main effect COND in time window IIbetween 130 and 240 ms: F1,20 = 4.26, P < 0.05; interactionCOND × CHAN: F1,50 = 2.68, P < 0.05). Moreover, for thedeviant condition, the ERP pattern is completed by a largeP3 wave (main effect COND in time window III between 240and 450 ms: F1,20 = 70.66, P < 0.001; interaction COND ×CHAN: F1,50 = 27.73, P < 0.001), peaking around 350 ms andshowing maximum amplitude at midline central andparietal sites.

There are obvious differences between the subject groupsin that the P3 appears to be considerably reduced for Chineseas compared to Western subjects. In the P3 window (timewindow III, 240–450 ms), the difference between standardsand deviants (i.e., the magnitude of the P3 effect) statisticallydepended on the subject group (interaction COND × GROUP:F1,20 = 5.33, P < 0.03), confirming the larger P3 effect inWesternas compared to Chinese musicians.

As shown by the topographies (see Fig. 2B), the P3components in both groups of musicians were prominent inparietal areas and without any hemispheric preponderance.

Furthermore, a three-way ANOVA analysis of the factorsCOND, CHAN, and SESSION (first or second exposure to thestimulus material) conducted within each musician grouprevealed a main “SESSION” effect in German musicians,F1,10 = 5.41, P < 0.05, which shows that German musiciansproduced larger P3 component in session 1 compared tosession 2. On the other hand, in session 1, the three-wayANOVA of the factors COND, CHAN, and GROUP evidenceda marginally significant GROUP effect (F1,9 = 3.54, P = 0.093)and a marginally significant interaction of COND × GROUP(F1,9 = 3.05, P = 0.115). No effects involving the GROUP factorwere found in session 2. No SESSION effect was found forChinese musicians, either.

2.2.2. Main experimentThe grand average ERP traces for the main experiment areillustrated in Fig. 3. Table 2 lists the results of the statisticalanalysis. Fig. 4 depicts the mean differences betweenphrased and unphrased items in the time windows between100 and 600 ms.

The traces in Fig. 3 show that for the unphrasedcondition, there is already a negative deflection starting

before the pause offset. This deflection somewhat mergeswith the N1 (especially for the Chinese music condition). Theremaining ERP pattern exhibits a series of negative and

185B R A I N R E S E A R C H 1 0 9 4 ( 2 0 0 6 ) 1 7 9 – 1 9 1

positive deflections peaking approximately 80 (negative), 200(positive), 400 (negative), and 550 ms (positive) after thepause offset. These peaks have different amplitude distribu-tions over electrode sites, phrasing conditions, subjectgroups, and musical styles. For analysis, time windowshave been chosen, each of which encompasses just onecomponent (see Experimental procedures and Fig. 3). Be-cause the present study focuses on the effects of theprocessing of phrase structure, only effects involving theCOND factor (comprising phrased/unphrased conditions) arereported. There were significant main effects for the CONDfactor for all the four time windows, reflecting the fact thatover the entire time period of analysis the amplitudes weremore positive going for the phrased as compared to unphrasedcondition.

The time window I, containing the N1, exhibits a maineffect of COND, suggesting that the N1 is more pronounced inthe unphrased condition. However, a closer look onto thewaveforms (Fig. 3) suggests that this negative shift could stemfrom another negative wave peaking around the pause offsetin the unphrased condition only. This component seems tooverlap with the N1.

For both types of music, the P2 is larger for Germanmusicians in the case of phrased melodies and larger forChinese musicians in the case of the unphrased versions, asconfirmed by a significant two-way interaction of COND ×GROUP for the average amplitude in time window II. Botheffects are considerably stronger for Chinese music (signifi-cant two-way interaction COND × STYLE and marginallysignificant (P = 0.055) three-way interaction of COND ×STYLE × GROUP; see also Figs. 3 and 4). No interactions withthe SESSION factor were found. Similar observations weremade for the time window III (Fig. 4, Table 2).

From Fig. 4, it seems that the CPS is larger for Germanas compared to Chinese musicians, as well as for Chineseas compared to Western music, without any interactionbetween the factors. However, while the influence of themusical style could be confirmed by statistics (COND ×STYLE interaction), no significant effect could be found forthe differences between the groups or sessions.

1 This in fact means that more Western pieces were erroneouslycategorized as Chinese, than the other way round, i.e., there is acertain inclination to categorize music as Chinese, if in doubt.2 Such a P3 reduction effect as a function of repetition has been

evidenced in previous studies (Debener et al., 2002; Ravden andPolich, 1998).

3. Discussion

Before discussing the electrophysiological effects of phrasingand culture, we first turn to the behavioral findings and theresults from our P3 tone perception test, in order tospecifically illuminate the differences between the subjectsgroups.

3.1. Behavioral data

In order to examine the influence of enculturation ontothe perception of music in musically trained subjects, itwould be ideal if each of the chosen subject groups isexpert in one and completely ignorant in the othermusical culture. Practically, however, the worldwide dom-inance of classical Western music renders compromisesinevitable. As our behavioral data clearly demonstrate, theChinese subjects recognized classical Western music as

reliably as the Germans did. Meanwhile, they are alsoquite good at recognizing traditional Chinese music (evenslightly better than in Western music1). German musi-cians, on the other hand, categorized a much largerportion of the Chinese pieces erroneously as Western.Hence, if it comes to categorizing music, there are nodifferences between the groups for Western music, whilefor Chinese music the Chinese subjects have a clearadvantage, although they did not receive any formaltraining in this type of music. This might be a result ofthe implicit exposure of the Chinese to traditional Chinese(besides classical and popular Western style) music duringtheir life before they left China.

The effect of phrasing onto the categorization perfor-mance, on the other hand, is equal for both groups ofmusicians. For both Chinese and Germans, phrasing seemsto facilitate categorization of their own type of music andhamper categorization of the respective culturally unfamiliarmusic type. A possible explanation of this phenomenon is thatphrasing causes some sort of “feeling of familiarity”, i.e.,phrased pieces are more easily categorized as music of one'sown culture.

3.2. The P3 experiment

The simple oddball paradigm using piano-like tones withoutany further musical context yielded a P3 like wave, whichwas larger for the German musicians. The topography of thecomponent as well as the paradigm of the experimentclearly indicates that it is a target P3 or P3b, reflectingattention allocation processes (Friedman et al., 2001; Levy etal., 2003). It has been shown that the target P3 amplitudecould be modulated by the arousal states of subjects due tonatural or environmental factors (for a detailed review, seePolich and Kok, 1995). More difficult tasks produce larger P3amplitudes due to the increased arousal states of thesubjects (Sommer et al., 1993) and so does the presentationof motivating instructions (Carrillo-de-la-Pena and Cada-veira, 2000). In the current study, the overall larger P3component for German musicians in the deviant condition ismainly originating from session 1 (first exposure to stimuli).Given the fact that the P3 experiment was conducted after atraining procedure for the main experiment, it is likely thatGerman musicians were relatively highly aroused by themore difficult task (unfamiliar Chinese music needed to becategorized) in the training process. This interpretation is inline with the findings that this effect is more obvious forsession 1 when the German subjects encounter the unfa-miliar Chinese music for the first time2. Results from themain experiment, i.e., the somewhat poorer behavioralperformance and the larger ERP activities for Germanmusicians compared to Chinese musicians in the Chinesemusic condition, support this interpretation.

186 B R A I N R E S E A R C H 1 0 9 4 ( 2 0 0 6 ) 1 7 9 – 1 9 1

Note that any anatomical explanations for the observeddifferences (e.g., in skull thickness or head shape) are notplausible because then they should have been equallyvisible in both sessions and all components, including theN1.

Most importantly, the fact that no SESSION effect could befound in the main experiment indicated that similar groupdifferences based on different arousal states did not play a rolethere.

3.3. The universality of the CPS component as a marker ofphrase perception

In the main experiment, we found a significant parietallydistributed positive deflection for the phrased compared to theunphrased listening condition in all four combinations ofsubject groups and musical styles. This component wasobserved between 450 ms and 600 ms post pause offset andclosely resembles the music CPS found in previous studies(Knösche et al., 2005; Neuhaus et al., 2006).

The CPS has been first found for the processing ofintonational phrase boundaries in language studies (Hruskaand Alter, 2004; Steinhauer, 2003; Steinhauer and Friederici,2001; Steinhauer et al., 1999), where it was elicited byprosodic cues and was interpreted to indicate the processingof intonational phrase boundaries. Compared to the CPSfound in language studies, the CPS in music (Knösche et al.,2005; Neuhaus et al., 2006) exhibits a different latency but asimilar amplitude and topography. The difference in latencyfor the CPS component between language and music studieswas interpreted by the different phrase boundary cues inlanguage compared to music, such as lengthened prefinalsyllables and changed F0 contours (Knösche et al., 2005). Thesimilar topography and mode of elicitation make it reason-able to draw the conclusion that both components mightreflect related cognitive processes.

The current study not only replicated the CPS finding forWestern music (Knösche et al., 2005) but demonstrated aCPS in a cross-cultural setting with a categorization task. Inthe work of Knösche et al. (2005), subjects were asked todetect rarely occurring wrong notes in the presented musicalpieces. This task was rather focusing the attention of thesubjects to the local structure of the music, while ourpresent categorization task required a more holistic ap-proach. The elicitation of the CPS was not fundamentallyinfluenced by this change of paradigm (see Fig. 3B ascompared to Knösche et al., 2005, please note the differentscales). This might be a hint that the music CPS reflectsprocesses that are at least partially concerned with phraseboundaries as structuring elements of the entire musicalpiece, rather than with the mere detection of the boundarymarkers, such as pauses (i.e., interruptions in a continuousstream of sound). There are also other reasons to believethat the CPS at least partially reflects higher cognitiveprocessing of phrase structure. First, the language CPS hasbeen shown to be elicited independently of the presence ofthe pause at the intonational phrase boundary (Steinhaueret al., 1999). Second, for the CPS in music phrasingperception, EEG and MEG based source localization resulthas shown that likely neural generators involve structures of

the limbic system associated with attention and memoryprocesses (Knösche et al., 2005). Third, the music CPS ismodulated by a number of factors other than pause, e.g.,musical experience and harmonic closure of the previousphrases (Neuhaus et al., 2006).

Furthermore, in our study, the presence of the CPSseems to be unaffected by the relation between theacculturation of the subjects and the cultural style of themusic (Fig. 4 and Table 2). Tentative influences of subjects'acculturation alone onto the amplitude of the CPS, assuggested by Fig. 4, could not be confirmed statistically(not even marginally) and must remain speculative, whilethe influence of the musical style could be proven.Independently of the cultural background of the listeners,Chinese music produced larger CPS amplitudes than West-ern music. The reason might be found in the properties ofthe stimulus material. Table 3 documents clear differencesbetween the two types of music: Chinese pieces have longerphrases, longer pauses, as well as longer notes immediatelybefore and after the pause. Therefore, this result is in linewith the observation by Neuhaus et al. (2006) that longerpauses as well as longer boundary tones produce larger CPSamplitudes.

An alternative explanation for the observed musical styleeffect lies in the fact that both groups of musiciansunderwent the same type of formal musical training basedon classical Western music. It could be that the CPS, incontrast to the earlier ERP components, reflects a processpredominantly influenced by formal musical training, ratherthan passive exposure to a musical environment. Thiswould be in line with the finding of Neuhaus et al. (2006)that explicit musical training has a profound effect onto theamplitude of the CPS.

Taken together, these results suggest that the CPS reflects amechanism, which (1) is related to the processing of phrasesand phrase boundaries, independently of the particular task ofthe subjects, (2) seems to reflect not only themere detection ofphrase boundary markers, but also the role of phraseboundaries as structuring elements of the entire musicalpiece, (3) is acquired or at least influenced by formal musicaltraining, and (4) is elicited by different music styles irrespec-tive of the relationship between the cultural background of thesubjects and the cultural style of the music.

3.4. Early effects of phrasing

The analysis of the time window I suggests that there mightbe a difference in N1 between phrased and unphrased items.Due to interference from ERP activity around the offset ofthe pause interval in the unphrased condition (possiblycaused by some auditory processing related to the pause-filling notes), an exact quantification of this effect is notpossible. It might, however, be related to the refractorinessmechanism of the neural system (Budd et al., 1998; Jones etal., 2000).

In the time range between 100 and 450 ms, the ERP did notonly react to the phrase boundary, but it was also affected bythe different musical styles and the cultural backgrounds ofthe subjects. Hence, the underlying neural processes aredriven by both stimulus features (i.e., bottom–up) and

187B R A I N R E S E A R C H 1 0 9 4 ( 2 0 0 6 ) 1 7 9 – 1 9 1

cultural-specific knowledge (i.e., top–down). Both types ofinformation interact during this processing phase, expressedby the statistical interactions between the factors COND(phrased or unphrased), STYLE (Chinese or Western music),and GROUP (Chinese or German musicians). As has beenpostulated by Narmour (1991), combined top–down andbottom–up systems are very important for predicting thecongruity of coming musical events when comparing themwith preceding contextual implications. Although both groupsof musicians have undergone the same formal training inWestern music (where they showed little difference bothbehaviorally and electrophysiologically), they differ in theirimplicit exposure to traditional Chinese music (giving theChinese a clear advantage with suchmusic, reflected in betterperformances). Such implicit cultural influence on musicprocessing has been demonstrated before in behavioralstudies (Drake and El Heni, 2003; Schellenberg and Trehub,1999).

4. Conclusion

The current study shows that the music CPS is also present,when subjects are required to perform a rather globalcategorization task instead of the more local wrong notedetection task used in (Knösche et al., 2005; Neuhaus et al.,2006). Moreover, the music CPS was observed in subjects ofdifferent cultural background listening both to music of theirnative and an alien culture. These findings add to thegenerality of the CPS as amarker for the processing of musicalphrasing. The results further suggest that the interactionbetween the acculturation of the listeners (top–down infor-mation) and the cultural style properties of the music(bottom–up information) influences phrase perception in arelatively early time window. In a recent behavioral study, itwas evidenced that musical structure organization is facili-tated, when listeners perceive music from their own culture(Drake and El Heni, 2003). The current ERP study adds to thedescription of this acculturation effect on music phrasing byspecifying its temporal characteristics.

Taken together, the results support the notion that,although musical phrase perception processes are basicallyuniversal (as the CPS effect is present in all conditions), theyare influenced by the relationship between musical style andacculturation of the subjects in a relatively early timewindow.

5. Experimental procedures

5.1. Subjects

Twenty-eight healthy trained musicians (14 German musi-cians, 14 Chinese musicians; all female) were paid forparticipation. None of the subjects had symptoms of anyneurological, psychiatric or internal disease nor had theytaken any medication on the day of examination and for atleast 3 days before. All subjects had normal hearing abilities.Due to EEG artefacts, two German and three Chinesesubjects were excluded from final analysis. Additionally,there was one pifa player among the recruited Chinese

musicians, whose data were deleted from the final analysisbecause this subject was the only one majored in Chineseinstrument playing. Thus, all information (including subjectinformation, behavioral data, and EEG data) mentionedhereafter refers to the remaining 12 German and 10 Chinesemusicians. All subjects were right-handed (average handed-ness: Chinesemusicians, 91.1 ± 5.97; Germanmusicians, 87.4 ±3.99) according to the Edinburgh Handedness Inventory (Oldfield,1971). The mean ages of the two group of subjects were botharound 24 years (German musicians, 24 ± 1.2 years; Chinesemusicians, 24.9 ± 1.4 years, respectively; age range 21–37, nosignificant difference between two musician groups). Theaverage starting age for instrument playing was about 7 yearsfor German musicians, about 8 years for Chinese musicians.Most of the subjects reported that they were able to play atleast two instruments (11 German musicians, 8 Chinesemusicians). The main present instrument for German musi-cians groupwere very diverse, including flute (5), accordion (1),violin (3), cello (1), piano (2); while for Chinese musicians, themain present instrument mostly focused on piano (7), alsosome violin (2), and keyboard (1). All subjects were currentlystill in the process of musical training, eighteen (11 Germanand 5 Chinese) had already passed their intermediate or finalexams. Including the training hours spent on classes, Germanmusicians reported on average to exercise 3.2 h per day,Chinese musicians 3.4 h per day. In addition to the dailyexercise, many musicians got chances to perform in public aswell (on average, German musicians: 5 times/year, Chinesemusicians: 4.2 times/year). Except for one Chinese musician,who trained for Jazz piano, all the other musicians from bothgroups reported to mainly play music from Baroque, Classicaland Romantic eras. This was the case for both training andspare time playing. Two of the Chinese subjects reported thatthey also preferred to play Chinese music in their spare time.Both groups of musicians reported a prominent Westernclassical hearing preference (German: 11.6 ± 3.4 h/week,Chinese: 10.9 ± 2.6 h/week), as well as the pop/rockmusic listening habits (German: 8.1 ± 1.8 h/week, Chinese:9.5 ± 3.1 h/week). All Chinese musicians had been brought upin China. In order to pursue a higher level of musicality inWestern classical music, they moved into the Westernmusical environment at the age of 18 to 24 years (on average21.2 ± 0.7 years) from China. Except for two of them, who usedto hear Chinese music occasionally in the past, none ofGerman musicians reported experience with Chinese music.Chinese musicians occasionally listened to Chinese musicafter they left China (on average 1.2 h/week). All subjects gavetheir informed consent prior to the first investigation.

5.2. Stimulus material and paradigm

In this study, synthesized melodies with piano-like timbrewere used. In order to explore any preexisting difference inauditory ERP components (e.g., due to anatomical differences)between the two groups of musicians, an oddball experimentwas conducted immediately after the short training sessionfor the main experiment. Three pairs of standard/devianttones were delivered successively within 3 blocks of theexperiment (120 standard and 30 deviant tones per block),each block containing one pair of stimuli, which were also

Table 3 – Comparison of phrase boundary featuresbetween Chinese and Western melodies

Chinesemelodies

Westernmelodies

Pause length (ms) 628 (39) 504 (25)Length of the preceding phrase (ms) 8355 (361) 5290 (209)Length of the last note before pause (ms) 765 (58) 511 (33)Length of the first note after pause (ms) 296 (21) 248 (26)

Values are averages, with 95% confidence intervals in parentheses.

188 B R A I N R E S E A R C H 1 0 9 4 ( 2 0 0 6 ) 1 7 9 – 1 9 1

synthesized piano-like tones (same timbre as in mainexperiment). Tones were 500 ms in duration. The inter-stimulus interval was randomized between 400 and 600 ms.The task of the subjects was to silently count the number of

Fig. 5 – (A) The sequence of events within one single trial. ISI inbiphrasal musical excerpts with different lengths ranging fromThere are 60 stimuli of each of the four stimulus types: phrasedWestern. Compared to the original (phrased) versions, the unphphrases with musically reasonable note(s).

deviant tones within each block and to give three final countsin the very end of the oddball experiment.

In the main experiment, we presented short melodicexcerpts (120 in total) belonging to Western music (60) orChinese traditional music (60). They were clearly structuredinto two phrases (each with 4 bars), the boundaries betweenwhich weremarked by pauses. In order to avoid excerpts fromthe musicians' daily practice repertoire, the melodies wereextracted from rather unknown musical pieces, with lengthsranging from 3000 to 17,000 ms.

For each of these melodies (labeled phrased), a modifiedcounterpart was created (the work was done by a profes-sional musician, who also evaluated all created examples),where the pauses were filled by one or several notes (labeledunphrased). An additional set of combined melodies was

dicates the interstimulus interval. The stimulus is one of the3000 to 17000 ms. (B) Typical biphrasal example stimuli.Chinese, unphrased Chinese, phrased Western, unphrasedrased ones were obtained by filling the pauses between two

189B R A I N R E S E A R C H 1 0 9 4 ( 2 0 0 6 ) 1 7 9 – 1 9 1

generated by combining two fractions (the cut-off pointswere set randomly along melodies) from different examples(but belonging to the same style and having the samemeter). These versions served as task items and wereexcluded from final analysis. A brief summary of theacoustic features of both groups of stimuli is listed inTable 3. Examples of stimuli are shown in Fig. 5B.

The experimental paradigm was similar to the previouswork of Knösche et al. (2005) and is sketched in Fig. 5A. Inorder to prevent any sequence effects, all the melodieswere presented in pseudo-random order. The resultingtrials were mixed up in different manners (balanced forsequence) and split into 6 blocks of equal length. About10% of the stimuli were task-relevant combined stimuli.Each experimental block was between 13 and 15 min longand contained an average of 20 Western and 20 Chinese aswell as 20 phrased and 20 unphrased stimuli, together with3 task items.

The current ERP experiment was carried out togetherwith a similar MEG experiment using the same subjectsand stimulus material (which is not reported due totechnical problems). In order to prevent any sequenceeffects, half of the subjects from each group first partici-pated in the current experiment, the other half in the MEGexperiment. There was at least a 10-day interval betweentwo sessions for each subject. The stimuli order wasbalanced between sessions. For the current experiment,this led to the fact that half of the data sets in each subjectgroup resulted from the first exposure to the stimulusmaterial, the other half from repeated exposure.

The task was to place each of the presented melodies intoone of the following 3 categories: “combined”, Chinese, orWestern. Before the real EEG measurement, a detailedinstruction and a training session were given to assure thatthe subjects understood the task well. The answers werecollected by delayed key-press (a visual prompt sentence wasgiven after the presentation of each melody, and also afeedback for the correctness of the answer was givenimmediately). They were not informed that the purpose ofthe experiment was the investigation of the perception ofphrasing in a cross-cultural context.

5.3. Recording procedure

The subjects were seated in the electrically shielded roomand fitted with an electrode cap with 51 Ag-AgCl electrodesconnected. Additional reference electrodes were placed atthe mastoid (left and right). Stimuli were presented vialoudspeaker, which was placed in a distance of about 1.5 m.Before starting the first block, loudness as well as centeringof the sound source was balanced individually. For gettingaccustomed to the stimuli, several examples were presentedbefore the start of the actual experiment. The subjects wereasked to avoid eye blinking, not to move, and to relax theirfacial muscles. They were also asked to look at a fixationpoint presented on the screen for the duration of eachmelody. Continuous EEG signals were recorded with abandpass filter of 0 to 100 Hz and digitized on-line with asampling rate of 250 Hz. Electro-oculograms (EOG) wererecorded from electrodes above and below the right eye and

at the outer canthi of the eyes. Impedances of electrodeswere kept below 5 kΩ. In an offline procedure, trials with eyeartefacts were marked and discarded. A high pass filter of0.25 Hz was applied offline in order to reduce very slowdrifts.

5.4. Signal analysis

Triggers were placed at the offsets of the phrase boundaries(the onset of the second phrase within each melody). Theaveraging window was set from −200 ms to 1000 ms relativeto the trigger points. In order to make sure that the signalused for baseline correction does not contain any signaldifference between the phrased and unphrased conditions, itwas chosen from the time interval −800 ms to −300 msrelative to the onset of the respective phrase boundary. Thatis relative to the offset of the last note in the 1st phrase.Averages of EEG recordings were computed for every subjectand condition. Trials with combined melodies (task items)were excluded from the averages. Sessions with more than50% contaminated trials were excluded altogether, resultingin the total exclusion of 2 German and 3 Chinese subjects.Finally, grand averages over subjects were computed forevery subject group (Germans and Chinese) and stimuluscondition (Chinese music, Western music, phrased,unphrased). Visual inspection of these grand averagestogether with the so-called running t tests (i.e., t testsperformed for each time step and channel without anymultiple testing correction) was used to identify time rangesof interest for the various ERP components. For the mainexperiment, the following time ranges were used foranalysis: (I): 40–100 ms (including the N1), (II): 100–300 ms(including the P2), (III): 300–450 ms, and (IV): 450–600 ms(including the CPS). For the P3 experiment, the analysis timewindows were (I): 0–130 ms (including the N1), (II): 130–240 ms (including the P2), and (III): 240–450 ms (including theP3). For each of the defined analysis time windows, potentialvalues were averaged over time samples and used asdependent variable of a four-way ANOVA with the within-subject factors “melodic condition” (COND, 2 levels: phrasedand unphrased), “channels” (CHAN, 51 levels) and “musicstyle” (STYLE, 2 levels: Western and Chinese), and thebetween-subject factor “cultural background” (GROUP, 2levels: German musicians and Chinese musicians). Additionallyto this four-way ANOVA, a five-way ANOVA, where theSESSION factor has been added, was also tested in order todetect any possible session effect. The P values of all effectswere corrected using the Greenhouse–Geisser method forrepeated-measure effects.

For the P3 experiment, a three-way ANOVA of the factorsCOND, CHAN, and GROUP was performed for each timewindow of interest. Moreover, for the time window III (P3component), a three-way ANOVA of the factors COND, CHAN,and SESSION was conducted within eachmusician group; anda three-way ANOVA of the factors COND, CHAN, and GROUPwas carried out within each session (session 1: first sessionwas EEG measurement; session 2: second session was EEGmeasurement).

Finally, in order to obtain a behavioral index of theinformation processing, the percentages of correct answers

190 B R A I N R E S E A R C H 1 0 9 4 ( 2 0 0 6 ) 1 7 9 – 1 9 1

were analyzed using a three-way ANOVA with the factorsCOND, GROUP, and STYLE.

Acknowledgments

The authors whish to thank Christiane Neuhaus for hervaluable help with the selection and preparation of thestimulation material as well as Cornelia Schmidt and HeikeBöthel for carrying out themeasurements. Thework describedin this paper was partially supported by a grant of the DeutscheForschungsgemeinschaft awarded to T.R.K. (KN 588/1-1).

R E F E R E N C E S

Arikan, M.K., Devrim, M., Oran, O., Inan, S., Elhil, M., Demiralp,T., 1999. Music effects on event-related potentials of humanson the basis of cultural environment. Neurosci. Lett. 268,21–24.

Balkwill, L.L., Thompson,W.F., 1999. A cross-cultural investigationof the perception of emotion in music: psychophysical andcultural cues. Music Percept. 17, 43–64.

Besson, M., 1998. Meaning, structure, and time in languageand music. Cah. Psychol. Cogn. Curr. Psychol. Cogn. 17,921–950.

Blacking, J., 1973. How Musical is Man? University of WashingtonPress, Seattle.

Budd, T.W., Barry, R.J., Gordon, E., Rennie, C., Michie, P.T., 1998.Decrement of the N1 auditory event-related potential withstimulus repetition: habituations vs. refractoriness. Int.J. Psychophysiol. 31, 51–68.

Carrillo-de-la-Pena, M.T., Cadaveira, F., 2000. The effect ofmotivational instructions on P300 amplitude. Neurophysiol.Clin.—Clin. Neurophysiol. 30, 232–239.

Castellano, M.A., Krumhansl, C.L., Bharucha, J.J., 1984. Tonalhierarchies in the music of north India. J. Exp. Psychol. Gen.113, 394–412.

Debener, S., Kranczioch, C., Herrmann, C.S., Engel, A.K., 2002.Auditory novelty oddball allows reliable distinction oftop–down and bottom–up processes of attention. Int.J. Psychophysiol. 46, 77–84.

Drake, C., Bertrand, D., 2001. The quest for universals in temporalprocessing in music. Ann. N.Y. Acad. Sci. 930, 17–27.

Drake, C., El Heni, J.B., 2003. Synchronizing with music:intercultural differences. Ann. N.Y. Acad. Sci. 999, 429–437.

Frankland, B.W., Cohen, A.J., 2004. Parsing of melody:quantification and testing of the local grouping rules of Lerdahland Jackendoff's a generative theory of tonal music. MusicPercept. 21, 499–543.

Friederici, A.D., Rueschemeyer, S.A., in press. Languageacquisition: biological versus cultural implications for brainstructure. In: Baltes Reuter-Lorenz, P.B., Rösler (Eds.), LifespanDevelopment and the Brain: The Perspective of BioculturalCo-Constructivism. Cambridge University Press, New York.

Friedman, D., Cycowicz, Y.M., Gaeta, H., 2001. The novelty P3: anevent-related brain potential (ERP) sign of the brain'sevaluation of novelty. Neurosci. Biobehav. Rev. 25, 355–373.

Gandour, J., Tong, Y.X., Wong, D., Talavage, T., Dzemidzic, M., Xu,Y.S., Li, X.J., Lowe, M., 2004. Hemispheric roles in the perceptionof speech prosody. NeuroImage 23, 344–357.

Genc, B.O., Genc, E., Tastekin, G., Ilhan, N., 2001.Musicogenic epilepsy with ictal single photon emissioncomputed tomography (SPECT): could these casescontribute to our knowledge of music processing? Eur.J. Neurol. 8, 191–194.

Hruska, C., Alter, K., 2004. How prosody can influence sentenceperception. In: Steube, A. (Ed.), Information Structure:Theoretical and Empirical Aspects, vol. 1. de Gruyter, Berlin,pp. 211–226.

Jacobsen, T., Schroger, E., Alter, K., 2004. Pre-attentive perceptionof vowel phonemes from variable speech stimuli.Psychophysiology 41, 654–659.

Jones, S.J., Vaz Pato, M., Sprague, L., 2000. Spectro-temporalanalysis of complex tones: two cortical processes dependenton the retention of sounds in the long auditory store. Clin.Neurophysiol. 111, 1569–1576.

Klein, D., Zatorre, R.J., Milner, B., Zhao, V., 2001. A cross-linguisticPET study of tone perception in mandarin Chinese and Englishspeakers. NeuroImage 13, 646–653.

Knösche, T.R., Neuhaus, C., Haueisen, J., Alter, K., Maess, B., Witte,O.W., Friederici, A.D., 2005. The perception of phrase structurein music. Hum. Brain Mapp. 24, 259–273.

Koelsch, S., Kasper, E., Sammler, D., Schulze, K., Gunter, T.C.,Friederici, A.D., 2004. Music, language and meaning: brainsignatures of semantic processing. Nat. Neurosci. 7, 302–307.

Krumhansl, C.L., 1995. Music psychology and music theory—Problems and prospects. Music Theory Spectr. 17, 53–80.

Krumhansl, C.L., 2000. Tonality induction: a statistical approachapplied cross-culturally. Music Percept. 17, 461–479.

Levy, D.A., Granot, R., Bentin, S., 2003. Neural sensitivity to humanvoices: ERP evidence of task and attentional influences.Psychophysiology 40, 291–305.

Lynch, M.P., Eilers, R.E., 1992. A study of perceptual developmentfor musical tuning. Percept. Psychophys. 52, 599–608.

Maess, B., Koelsch, S., Gunter, T.C., Friederici, A.D., 2001. Musicalsyntax is processed in Broca's area: an MEG study. Nat.Neurosci. 4, 540–545.

Morrison, S.J., Demorest, S.M., Aylward, E.H., Cramer, S.C.,Maravilla, K.R., 2003. fMRI investigation of cross-cultural musiccomprehension. NeuroImage 20, 378–384.

Narmour, E., 1991. The top–down and bottom–up systems ofmusical implication—Building on Meyer theory of emotionalsyntax. Music Percept. 9, 1–26.

Neuhaus, C., 2003. Perceiving musical scale structures: across-cultural event-related brain potentials study. Ann. N.Y.Acad. Sci. 999, 184–188.

Neuhaus, C., Knösche, T.R., Friederici, A.D., 2006. Effects ofmusicalexpertise and boundary markers on phrase perception inmusic. J. Cogn. Neurosci. 18, 1–22.

Oldfield, R.C., 1971. The assessment and analysis of handedness.Neuropsychologia 9, 97–113.

Pannekamp, A., Toepel, U., Alter, K., Hahne, A., Friederici, A.D.,2005. Prosody driven sentence processing: an event-relatedbrain potential study. J. Cogn. Neurosci. 17, 407–421.

Patel, A.D., 2003. Rhythm in language and music—Parallels anddifferences. Ann. N.Y. Acad. Sci. 999, 140–143.

Patel, A.D., Peretz, I., Tramo, M., Labreque, R., 1998. Processingprosodic and musical patterns: a neuropsychologicalinvestigation. Brain Lang. 61, 123–144.

Polich, J., Kok, A., 1995. Cognitive and biological determinants ofP300: an integrative review. Biol. Psychol. 41, 103–146.

Ravden, D., Polich, J., 1998. Habituation of P300 from visual stimuli.Int. J. Psychophysiol. 30, 359–365.

Schellenberg, E.G., Trehub, S.E., 1999. Culture-general andculture-specific factors in the discrimination of melodies.J. Exp. Child Psychol. 74, 107–127.

Schlosser, M.J., Aoyagi, N., Fulbright, R.K., Gore, J.C., McCarthy, G.,1998. Functional MRI studies of auditory comprehension. Hum.Brain Mapp. 6, 1–13.

Smith, L.D., Williams, R.N., 1999. Children's artistic responses tomusical intervals. Am. J. Psychol. 112, 383–410.

Sommer, W., Leuthold, H., Hermanutz, M., 1993. Covert effects ofalcohol revealed by event-related potentials. Percept.Psychophys. 54, 127–135.

191B R A I N R E S E A R C H 1 0 9 4 ( 2 0 0 6 ) 1 7 9 – 1 9 1

Steinhauer, K., 2003. Electrophysiological correlates of prosodyand punctuation. Brain Lang. 86, 142–164.

Steinhauer, K., Friederici, A.D., 2001. Prosodic boundaries, commarules, and brain responses: the closure positive shift in ERPs asa universal marker for prosodic phrasing in listeners andreaders. J. Psycholinguist. Res. 30, 267–295.

Steinhauer, K., Alter, K., Friederici, A.D., 1999. Brain potentialsindicate immediate use of prosodic cues in natural speechprocessing. Nat. Neurosci. 2, 191–196.

Warren, P., Grabe, E., Nolan, F., 1995. Prosody, phonology andparsing in closure ambiguities. Lang. Cogn. Processes 10,457–486.